Building natural language generation systems

Review: This book is probably the first of its kind. Although there has been considerable work in natural language summarization, research in producing natural language is in its infancy. The authors draw upon actual NLG systems to illustrate the techinical issues involved. They are careful to point out that NLG may not be the best solution for the reader's practical problems. For instance, human writers might produce better text, or mail-merge programs might do the job (albeit with lower quality) at a significantly lower price. A shortcoming -- perhaps a necessary one -- is that the authors hesitate to describe research and development trends for the past decade. The reason does not appear to be a shortage of the quality and quantity of NLG R&D, because the authors do analyze trends for the 1970's and the 1980's. Are the authors hesitant to criticize their contemporary colleagues? A blunt trends analysis would have been helpful, particularly for readers new to the field.